[gmx-users] Paralell problems...

[Florian Haberl]

hi,

On Thursday 01 June 2006 21:43, Jorge Hernandez Fernandez wrote:
> Dear GRMxers:
>
>         We need some help in running our Gromacs 3.2.1 parallel environment
> in 5 nodes of SunFires V20v (dual-core Opteron 2.2, 4 GB DDR2 RAM), gigabit
> eth, mpich-1.2.7, using Sun Grid Engine 6.0.

Update to gromacs 3.3.1, with normal system sizes gromacs run well up to 4 
cpus here (dual boards). Else update your interconnection type.

> In a 2 processors job (single node), the results are O.K. but in a 4, 6 or
> 8 processors (2, 3 or 4 nodes) in the machine log files we have:
>
> In “ -np 4” the error was :
>
> p4_5991:  p4_error: Timeout in establishing connection to remote process: 0
> p0_8456:  p4_error: net_recv read:  probable EOF on socket: 1
> p5_5996:  p4_error: Timeout in establishing connection to remote process: 0
> p0_8456: (333.761719) net_send: could not write to fd=4, errno = 32
>
> In  “ –np 8” the error was:
>
> p0_16398:  p4_error: interrupt SIGSEGV: 11
>
> In all the cases, we obtain a “Killed by signal 2” at the end
> of the  gromacs log,  and process stopped.
> Grompp was executed with the -np option, and our final script was:
>
> #$ -v MPIRUN
> #$ -v MPICH_PROCESS_GROUP
>
> MPIRUN -np $NSLOTS -machinefile $TMPDIR/machinefile /nfs/gromacs-3.2.
> 1_paralelo/x86_64-unknown-linux-gnu/bin/mdrun -np 6 -s ABint/ABint_md.tpr
> -o ABint/ABint_md.trr -c ABint/ABint_md.gro -e ABint/ABintener.edr -g
> ABint/ ABintmd.log -nice 0
>
> Any help will be extremely appreciated:
>                                                    Jorge H.F.
>
> --
> == Dr. JORGE HERNANDEZ FERNANDEZ ==
>
> ====== Center of Applied Toxinology ======
> ===== CAT-CEPID - Instituto Butantan =====
> ======  Ave Vital Brasil, 1500 S.P. ======
> Tel: 055 11 3726 7222 r.2042 Fax: 055 11 3721 6605
>
> ==== S.B.I.- EMBRAPA - BioInformatica ====
> C.p. 6041 Cidade Universitária "Zeferino Vaz"
> ===== Barão Geraldo Campinas S.P. ========
> =========== 13080-970 ====================
> Tel:  055 19-37895828  Cell: 055 11-97126104

Greetings,

Florian

-- 
-------------------------------------------------------------------------------
 Florian Haberl                        
 Computer-Chemie-Centrum   
 Universitaet Erlangen/ Nuernberg
 Naegelsbachstr 25
 D-91052 Erlangen
 Mailto: florian.haberl AT chemie.uni-erlangen.de
-------------------------------------------------------------------------------